计算机应用研究2018，Vol.35Issue(1)：88-92,5.DOI:10.3969/j.issn.1001-3695.2018.01.017
一种双重特征选择的不平衡复杂网络链接分类模型
Dual feature selection imbalanced complex network link classification model
摘要
Abstract
Links classification based on supervised classification is a main research area in the field of complex network analysis,the core idea of link prediction is that the network is divided into training and testing networks,then it forecasts testing sample by learning and training the training set.However,in this scenario,the distribution of positive samples and negative samples with different category is unbalanced,there also will be redundancy between features,and such phenomenon often restricts the classification performance.For this problem,this paper proposed a dual feature selection classification model,such model used K-means clustering algorithm for unbalanced feature to solve the problem of data imbalance.And it also incorporated Relief for assigning weights to features and minimum redundancy-maximum relevance (mRMR) to measure the correlation between the characteristics and features and characteristics and between categories.Experimental results on several real complex network datasets show that,when comparing to the current links classification model,the proposed method can significantly improve the classification performance.关键词
链接分类/Relief/K-均值/特征选择/mRMR/不平衡问题Key words
link classification/Relief/K-means/feature selection/mRMR/imbalance problem分类
信息技术与安全科学引用本文复制引用
伍杰华,徐宏..一种双重特征选择的不平衡复杂网络链接分类模型[J].计算机应用研究,2018,35(1):88-92,5.基金项目
广东省优秀青年教师资助项目(YQ2015177) （YQ2015177）
广东省科技计划资助项目(2017ZC0303) （2017ZC0303）
